Hey, it’s Marc!
AI agents are getting smarter — and harder to control.
That’s why a new crop of “guardian agents” is emerging to monitor and manage them. By 2030, they could own 15% of the AI agent market, given 70% of AI applications will use multi-agent systems by 2028. [Read more]
The bigger question isn't whether we need AI to watch AI, but whether we're building the guardrails fast enough.
Here are the top Agent signals:
Databricks launches Agent Bricks, a new tool that lets enterprises build production-grade AI agents using their own data — with zero manual tuning
ThinkDeep AI launches DeepBrain, a sovereign multi-agent AI platform used by the French government to automate public services like fraud detection, legal document analysis, and citizen inquiries
Zip launches 50 AI agents to automate procurement workflows
Indicium launches AI Data Squads, a new consulting service that combines human experts with AI agents to modernise and migrate legacy data systems, starting with platforms like Databricks and dbt Labs
👉 Get your brand in front of 30,000+ decision-makers — book your ad spot now.
Top Boardroom Reads This Week
The Gentle Singularity (Sam Altman)
MCP: A Strategic Foundation for Enterprise-Ready AI Agents (CIO)
Are we ready to hand AI agents the keys? (MIT Tech Review)
HBR study: What are people really using GenAI for in 2025? (51)
Apple vs OpenAI (51)
Sovereign AI Agents Think Local, Act Global With NVIDIA AI Factories (Nvidia)
Artificial Intelligence: Trends 2025 (Bond Capital)
The state of AI 2025 (51)
Orchestrating agentic AI for intelligent business operations (IBM)
State-Of-The-Art Prompting For AI Agents (YCombinator)
Humanity Deserves Better (Jony Ive and Laurene Powell)
How Generative Engine Optimization (GEO) Rewrites the Rules of Search (a16z)
AI Agent versus MCP (ByteByteGo)
Sundar Pichai on AI, Antitrust Suits, Privacy, Google Search, Gadgets (Bloomberg)
Power Moves
NVIDIA is building $200B sovereign AI factories (to break hyperscaler lock-in)
NVIDIA just unveiled Enterprise AI Factory designs targeting Europe's $200B AI investment—complete infrastructure stacks that let countries and enterprises build AI on-premises instead of relying on AWS/Azure/GCP. The package includes Blackwell chips, NIM microservices supporting 100,000+ models from Hugging Face, and AI safety blueprints. [ANNOUNCEMENT]
The strategic play: While hyperscalers want you to rent their AI in the cloud, NVIDIA is selling the factory itself. Major European enterprises like BNP Paribas, L'Oreal, and BT Group are already building NVIDIA-powered AI factories for financial services, beauty matching, and network optimisation.
So what: The AI infrastructure war isn't just about chips—it's about control. Sovereign AI factories mean you can train models on sensitive data without sending it to Big Tech clouds. When regulators start demanding data residency for AI workloads, on-premises factories become a competitive advantage.
Apple just killed the AI reasoning hype
Apple researchers tested OpenAI's o1, Claude, and DeepSeek's "reasoning" models on puzzle tests and found they don't actually reason—they memorize patterns and give up when tasks get complex. The kicker: simple LLMs often outperform expensive "reasoning" models on low-complexity tasks, while both collapse on high-complexity problems. So what? f you're deploying AI for complex forecasting, risk analysis, or multi-step reasoning, you're betting on capabilities that don't exist yet. Meta's Yann LeCun agrees—calling current LRMs "glorified autocomplete engines with good marketing." [See analysis below]
Vendor Spotlight: Manus
Your C-suite just asked you to do more with less. Again.
Meet Manus AI.
Launched earlier this year by Chinese startup Monica AI, Manus is the first truly autonomous AI agent that handles complex, multi-step business tasks — end-to-end.
Recently, it introduced its “Chat Mode,” free and unlimited for all users.
Why it matters: Most AI tools need babysitting. Manus doesn’t.
Assign a project → it splits tasks across agents
Hooks into your tools (browsers, databases, APIs)
Works while you're offline
Learns and adapts to your workflow
What businesses are using it for
Across functions, Manus is a full-stack workhorse:
Recruiting → resume screening, candidate insights
Finance → reports, market analysis
Ops → inbox parsing, meeting recaps
Marketing → SEO, content, campaign strategy, deck design
R&D → competitor research, technical summaries
Creative → now with Veo3 for AI video creation
Others → deck design, website creation, data storytelling
Traction
138K+ users in days
GAIA benchmark leader (>75%)
It has also partnered with Alibaba's Qwen AI, which gives Manus:
Enterprise-grade security
Massive computational power
24/7 global reliability
Same infrastructure running Alibaba's $500B+ business
If your GenAI tools still need constant prompting — they’re not agents.
Manus is.
More on Manus AI:
Top AI Vendors This Week
👉 We’re tracking 1000s of AI and blockchain vendors so you find the right vendors, partners, and bet on the right tech. Sign up for early access.
More from 51:
That’s all for today.
Thanks,
Marc & Team